Fault detection and classification of metal nuts containing anomaly by deep learning techniques

Rapid changes and developments in information and communication systems have caused a sudden paradigm shift in every field connected with technology. This paradigm shift made its presence felt simultaneously in the industrial field, has found its place in accordance with the needs of the sector and embodied as Industry 4.0 within the sector. This paradigm, known as the 4th Industrial Revolution, with the help of concept of the IoT, has implemented modern automation systems, production technologies and intelligent industrial systems, which are in instant communication with each other. In this smart communication network, reporting, analysis and cost-effective management of shared data and development of efficient business models are of great importance for enterprises. This process requires industrial enterprises to transform their business processes, into modern automation and production units that are compatible with the future by integrating business processes with intelligent systems. In this sense, computer vision applications are the leading solutions that capture the age and extend to the future. Businesses that have gone through automation and production processes have begun to integrate techniques such as machine learning into their specific operations. The production process is particularly important in this integration since it forms a joint point with the main headings of industrial enterprises such as raw material usage, quality control and operational efficiency. This study proposes an intelligent approach used to detect faulty products and optimize production efficiency in the industrial area. The aim of this study is to develop a business intelligence application that detects and separates the defective products that occur during production, helps to determine the cause of the error, contributes to the reduction of the error rate and increase the quality, optimizes the use of raw materials, increases production efficiency and ensures low cost operation of the process by replacing the human factor with computer vision in industrial enterprises producing screws and nuts. Software and frameworks used to design this computer vision application are Python, OpenCV, Tensorflow Object Detection API, Tensorboard, GoogleColab, and Visual Studio Code. Keywords: Anomaly, Fault, Computer Vision, Deep Learning, Object Detection, Classification